com.bigdata.btree.IRangeQuery Maven / Gradle / Ivy
/**
Copyright (C) SYSTAP, LLC DBA Blazegraph 2006-2016. All rights reserved.
Contact:
SYSTAP, LLC DBA Blazegraph
2501 Calvert ST NW #106
Washington, DC 20008
[email protected]
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
/*
* Created on Feb 14, 2007
*/
package com.bigdata.btree;
import java.util.Iterator;
import com.bigdata.bfs.BigdataFileSystem;
import com.bigdata.btree.IndexSegment.IndexSegmentTupleCursor;
import com.bigdata.btree.filter.ITupleFilter;
import com.bigdata.btree.filter.TupleRemover;
import com.bigdata.service.IBigdataClient;
import com.bigdata.service.IDataService;
import com.bigdata.service.ndx.IClientIndex;
import cutthecrap.utils.striterators.IFilter;
import cutthecrap.utils.striterators.Striterator;
/**
* Interface for range count and range query operations.
*
* Note: There are implementations of this interface for both local indices and
* for distributed scale-out indices.
*
* @author Bryan Thompson
* @version $Id$
*/
public interface IRangeQuery {
/**
* Return the #of tuples in the index.
*
* Note: If the index supports deletion markers then the range count will be
* an upper bound and may double count tuples which have been overwritten,
* including the special case where the overwrite is a delete.
*
* @return The #of tuples in the index.
*
* @see ICheckpointProtocol#rangeCount()
*/
public long rangeCount();
/**
* Return the #of tuples in a half-open key range. The fromKey and toKey
* need not exist in the B+Tree.
*
* Note: If the index supports deletion markers then the range count will be
* an upper bound and may double count tuples which have been overwritten,
* including the special case where the overwrite is a delete.
*
* @param fromKey
* The lowest key that will be counted (inclusive). When
* null there is no lower bound.
* @param toKey
* The first key that will not be counted (exclusive). When
* null there is no upper bound.
*
* @return The #of tuples in the half-open key range.
*
* @todo ??? throw exception if LT but allow GTE? This will be zero if
* toKey is less than or equal to fromKey in the total
* ordering.
*/
public long rangeCount(byte[] fromKey, byte[] toKey);
/**
* Return the exact #of tuples in a half-open key range. The fromKey and
* toKey need not exist in the B+Tree.
*
* Note: If the index supports deletion markers then this operation will
* require a key-range scan.
*
* @param fromKey
* The lowest key that will be counted (inclusive). When
* null there is no lower bound.
* @param toKey
* The first key that will not be counted (exclusive). When
* null there is no upper bound.
*
* @return The exact #of tuples in the half-open key range.
*/
public long rangeCountExact(byte[] fromKey, byte[] toKey);
/**
* Return the exact #of tuples in a half-open key range, including any
* deleted tuples. The fromKey and toKey need not exist in the B+Tree.
*
* When the view is just an {@link AbstractBTree} the result is the same as
* for {@link IRangeQuery#rangeCount(byte[], byte[])}, which already
* reports all tuples regardless of whether or not they are deleted.
*
* When the index is a view with multiple sources, this operation requires a
* key-range scan where both deleted and undeleted tuples are visited.
*
* @param fromKey
* The lowest key that will be counted (inclusive). When
* null there is no lower bound.
* @param toKey
* The first key that will not be counted (exclusive). When
* null there is no upper bound.
*
* @return The exact #of deleted and undeleted tuples in the half-open key
* range.
*
* @see IRangeQuery#rangeCountExact(byte[], byte[])
*/
public long rangeCountExactWithDeleted(final byte[] fromKey,
final byte[] toKey);
/**
* Flag specifies no data (the #of scanned index entries matching the optional
* filter will still be reported).
*/
public static final int NONE = 0;
/**
* Flag specifies that keys in the key range will be returned. The keys are
* guaranteed to be made available via {@link ITupleIterator#getKey()} only
* when this flag is given.
*/
public static final int KEYS = 1 << 0;
/**
* Flag specifies that values in the key range will be returned. The values
* are guaranteed to be made available via {@link ITupleIterator#next()} and
* {@link ITupleIterator#getValue()} only when this flag is given.
*/
public static final int VALS = 1 << 1;
/**
* Flag specifies that deleted index entries for a key are visited by the
* iterator (by default the iterator will hide deleted index entries).
*/
public static final int DELETED = 1 << 2;
/**
* The flags that should be used by default [{@link #KEYS}, {@link #VALS}]
* in contexts where the flags are not explicitly specified by the
* application such as {@link #rangeIterator(byte[], byte[])}.
*/
public static final int DEFAULT = KEYS | VALS;
/**
* Shorthand for [{@link #KEYS}, {@link #VALS}, {@link #DELETED}].
*/
public static final int ALL = (KEYS | VALS | DELETED);
/**
* Flag specifies that the iterator, including any {@link ITupleFilter}s,
* will not write on the index. Various optimizations may be applied when
* this flag is present. (Read only can be inferred if {@link #CURSOR} flag
* is NOT specified AND there are NO {@link ITupleFilter}s).
*/
public static final int READONLY = 1 << 3;
/**
* Flag specifies that entries visited by the iterator in the key range will
* be removed from the index. This flag may be combined with
* {@link #KEYS} or {@link #VALS} in order to return the keys and/or values
* for the deleted entries. When a {@link IFilter} is specified, the filter
* stack will be applied first and then {@link #REMOVEALL} will cause a
* {@link TupleRemover} to be layered on top. You can achieve other stacked
* iterator semantics using {@link IFilter}, including causing
* {@link ITuple}s to be removed at a different layer in the stack. Note
* however, that removal for a local {@link BTree} will require that the
* {@link TupleRemover} is stacked directly over an {@link ITupleCursor}.
*
* Note: This semantics of this flag require that the entries are atomically
* removed within the isolation level of the operation. In particular, if
* the iterator is running against an {@link IDataService} using an
* unisolated view then the entries MUST be buffered and removed as the
* {@link ResultSet} is populated.
*
* Note: The {@link BigdataFileSystem#deleteHead(String, int)} relies on
* this atomic guarantee.
*
* @todo define rangeRemove(fromKey,toKey,filter)? This method would return
* the #of items matching the optional filter that were deleted. It
* will be a parallelizable operation since it does not specify a
* limit on the #of items to be removed and does not return any data
* or metadata (other than the count) for the deleted items.
*
* Note: We still need {@link #REMOVEALL} since it provides an atomic
* remove with return of an optionally limited #of matching index
* entries. This makes it ideal for creating certain kinds of queue
* constructions.
*/
public static final int REMOVEALL = 1 << 4;
/**
* Flag specifies that the base iterator will support the full
* {@link ITupleCursor} API, including traversal with concurrent
* modification, bi-directional tuple navigation and random seeks within the
* key range. There are several pragmatic reasons why you would or would not
* specify this flag.
*
* - The original {@link Striterator} construction for the {@link BTree}
* is faster than the newer {@link AbstractBTreeTupleCursor}. It
* is used by default when this flag is NOT specified and the iterator is
* running across a {@link BTree}. (The {@link IndexSegmentTupleCursor} is
* used for {@link IndexSegment}s regardless of the value of this flag
* since it exploits the double-linked leaves of the {@link IndexSegment}
* and is therefore MORE efficient than the {@link Striterator} based
* construct.)
* - This flag enables traversal with concurrent modification (i.e.,
* {@link Iterator#remove()}) when used with a local {@link BTree}.
* Scale-out iterators always support traversal with concurrent modification
* since they heavily buffer the iterator with {@link ResultSet}s.
*
*/
public static final int CURSOR = 1 << 5;
/**
* Flag specifies that the entries will be visited using a reverse scan. The
* first tuple to be visited will be the tuple having the largest key
* strictly less than the optional upper bound for the key range. The
* iterator will then visit the previous tuple(s) until it has visited the
* tuple having the smallest key greater than or equal to the optional lower
* bound for the key range.
*
* This flag may be used to realize a number of interesting constructions,
* including atomic operations on the tail of a queue and obtaining the last
* key in the key range.
*/
public static final int REVERSE = 1 << 6;
/**
* There are two ways in which the successor of an unsigned byte[] key may
* be computed. The default is to append an unsigned zero byte to the key.
* This forms the next possible key for the natural unsigned byte[] sort
* order.
*
* The other choice is to treat the unsigned byte[] as a fixed length bit
* string and to add ONE (1) with rollover. This works in some special
* circumstances, primarily because you are actually seeking to skip over
* all keys that have a given key as their prefix and continue the iterator
* at the next possible prefix having the same length.
*
* Note: This option is applied by {@link AbstractChunkedTupleIterator},
* which is responsible for issuing continuation queries.
*/
public static final int FIXED_LENGTH_SUCCESSOR = 1 << 7;
/**
* Flag indicates that the iterator may process multiple index partitions in
* parallel. While tuples are visited in each index partition in the natural
* ordering, this flag breaks the total ordering guarantee of the iterator
* if multiple index partitions are visited as tuples from each index
* partition will be visited concurrently and appear in an interleaved
* ordering. To maintain efficiency, the tuples are interleaved in chunks so
* processing which benefits from order effects can exploit the within chunk
* ordering of the tuples. The semantics of this flag are realized by the
* {@link IClientIndex} implementation. This flag has no effect on an
* {@link ILocalBTreeView} and is not passed through to the
* {@link IDataService}.
*
* @todo This flag is not supported in combination with {@link #CURSOR}?
* @todo This flag is not supported in combination with {@link #REVERSE}?
*/
public static final int PARALLEL = 1 << 8;
/**
* Visits all tuples in key order. This is identical to
*
*
* rangeIterator(null, null)
*
*
* @return An iterator that will visit all entries in key order.
*/
public ITupleIterator rangeIterator();
/**
* Return an iterator that visits the entries in a half-open key range. When
* toKey EQ fromKey nothing will be visited. It
* is an error if toKey LT fromKey.
*
* @param fromKey
* The first key that will be visited (inclusive lower bound).
* When null there is no lower bound.
* @param toKey
* The first key that will NOT be visited (exclusive upper
* bound). When null there is no upper bound.
*
* @throws RuntimeException
* if fromKey is non-null and orders LT
* the inclusive lower bound for an index partition.
*
* @throws RuntimeException
* if toKey is non-null and orders GTE
* the exclusive upper bound for an index partition.
*
* @see SuccessorUtil, which may be used to compute the successor of a value
* before encoding it as a component of a key.
*
* @see BytesUtil#successor(byte[]), which may be used to compute the
* successor of an encoded key.
*
* @see EntryFilter, which may be used to filter the entries visited by the
* iterator.
*
* @todo define behavior when the toKey is less than the fromKey.
*/
public ITupleIterator rangeIterator(byte[] fromKey, byte[] toKey);
/**
* Designated variant (the one that gets overridden) for an iterator that
* visits the entries in a half-open key range. When toKey
* EQ fromKey nothing will be visited. It is an error if
* toKey LT fromKey.
*
* @param fromKey
* The first key that will be visited (inclusive lower bound).
* When null there is no lower bound.
* @param toKey
* The first key that will NOT be visited (exclusive upper
* bound). When null there is no upper bound.
* @param capacity
* The #of entries to buffer at a time. This is a hint and MAY be
* zero (0) to use an implementation specific default
* capacity. A non-zero value may be used if you know that you
* want at most N results or if you want to override the default
* #of results to be buffered before sending them across a
* network interface. (Note that you can control the default
* value using
* {@link IBigdataClient.Options#DEFAULT_CLIENT_RANGE_QUERY_CAPACITY}).
* @param flags
* A bitwise OR of {@link #KEYS}, {@link #VALS}, etc.
* @param filterCtor
* An optional object used to construct a stacked iterator. When
* {@link #CURSOR} is specified in flags, the base
* iterator will implement {@link ITupleCursor} and the first
* filter in the stack can safely cast the source iterator to an
* {@link ITupleCursor}. If the outermost filter in the stack
* does not implement {@link ITupleIterator}, then it will be
* wrapped an {@link ITupleIterator}.
*
* @see SuccessorUtil, which may be used to compute the successor of a value
* before encoding it as a component of a key.
*
* @see BytesUtil#successor(byte[]), which may be used to compute the
* successor of an encoded key.
*
* @see IFilterConstructor, which may be used to construct an iterator stack
* performing filtering or other operations.
*/
public ITupleIterator rangeIterator(byte[] fromKey, byte[] toKey,
int capacity, int flags, IFilter filterCtor);
// /**
// * An iterator that is mapped over a set of key ranges.
// *
// * @param fromKeys
// * An array of inclusive lower bounds with one entry for each key
// * range over which the iterator will be mapped. The elements of
// * the array MUST be sorted. The first element MAY be a
// * null to indicate that there is no lower bound
// * for the first key range.
// * @param toKeys
// * An array of exclusive upper bounds with one entry for each key
// * range over which the iterator will be mapped. Each toKey MUST
// * be GTE the corresponding fromKey. The last element MAY be a
// * null to indicate that there is no exclusive
// * upper bound bound for the last key range.
// * @param capacity
// * The #of entries to buffer at a time. This is a hint and MAY be
// * zero (0) to use an implementation specific default
// * capacity. The capacity is intended to limit the burden on the
// * heap imposed by the iterator if it needs to buffer data, e.g.,
// * before sending it across a network interface.
// * @param flags
// * A bitwise OR of {@link #KEYS}, {@link #VALS}, etc.
// */
// public ITupleIterator rangeIterator(byte[][] fromKeys, byte[][] toKeys,
// int capacity, int flags);
// removeAll() could be added, but the problem is that we often want
// the keys or values of the deleted entries, at which point you have to
// use the rangeIterator anyway.
// /**
// * Removes all entries in the key range from the index. When running on a
// * scale-out index, this operation is atomic for each index partition. The
// * operation may be used to build queue-like constructed by atomic delete of
// * the first item in a key range. This operation is parallelized across
// * index partitions when no limit is specified and serialized across index
// * partitions when a limit is specified.
// *
// * @param fromKey
// * The first key that will be visited (inclusive). When
// * null there is no lower bound.
// * @param toKey
// * The first key that will NOT be visited (exclusive). When
// * null there is no upper bound.
// * @param limit
// * When non-zero, this is the maximum #of entries that will be
// * removed. When zero (0L) all index entries in the key range
// * will be removed.
// *
// * @return The #of index entries that were removed.
// */
// public long removeAll(byte[] fromKey,byte[] toKey, long limit);
}